Network cameras have been developed that can continuously monitor a scene and store long sequences of recorded video for later review either locally at the camera or using a recording device such as a digital video recorder (DVR) or network video recorder (NVR) or remotely using cloud-based storage. Some systems having network cameras can detect moving objects in the recorded video stream and can generate a synopsis video of the detected moving objects by providing portions of the recorded video stream in a new synopsis video.
However, these conventional systems have some limitations that can prevent widespread implementation and consumer interest. For example, because the original recorded video stream is so large, local processing servers are often necessary to generate synopsis videos. These systems can be expensive and difficult to install and can therefore limit the interest of a typical consumer. Moreover, even with the necessary equipment, in conventional systems, it can be difficult for a user to sort through and select objects of interest from a crowded synopsis video that includes overlapping objects selected from a long and continuous input video stream.
There is therefore a need in the art for improved systems and methods for generating synopsis videos.